Repository navigation
chore: [DO NOT MERGE] CI run for rebased beta - #3111
Conversation
Targets the recently released Appium 4 beta (4.0.0-beta.1). Bumps the appium peer dependency and all lockstep @appium/* packages to their Appium 4 beta releases, and updates driver code for the resulting API removals/changes described in the Appium 3-to-4 migration guide (appium/appium#22790). - @appium/strongbox, @appium/types, @appium/docutils, @appium/oxc-config, @appium/semantic-release-config, @appium/tsconfig bumped to their Appium 4 lockstep beta releases (engine/ESM-only bumps, verified byte-identical source vs. master where diffable, not API changes) - Removed the driverData mechanism and the legacy 4-argument createSession override (both removed upstream); createSession now takes the single w3cCapabilities argument - Removed the already-deprecated reset API (driver.reset() / POST /session/:sessionId/appium/app/reset), which only ever threw a deprecation error - Fixed a real regression in the W3C timeouts command: BaseDriver's generic timeouts() handler no longer dispatches through scriptTimeoutW3C/pageLoadTimeoutW3C/implicitWaitW3C - it now calls setScriptTimeout/setPageLoadTimeout/setImplicitWait directly for both the legacy and W3C timeout forms. Renamed setAsyncScriptTimeout to setScriptTimeout to match, dropped the now-unreachable W3C/MJSONWP wrapper methods, and made sure BaseDriver's own scriptTimeoutMs/ pageLoadTimeoutMs stay in sync too, since GET /timeouts now reports them on Appium 4 (it never did on Appium 3) - Replaced the Location type (removed from @appium/types along with the legacy geolocation endpoint) with a local LocationWithAltitude interface - Narrowed fs.glob()'s now-widened overloaded return type at the handful of call sites that don't use its withFileTypes/lazy options - Made getStrings's language parameter optional in its type signature, matching its actual (always-optional) runtime/wire behavior - caught by the new compile-time execute-method-map param checking - Bridged two known upstream typing lags (appium-webdriveragent and appium-ios-simulator still resolve AppiumLogger from pre-Appium4 @appium/types copies) with documented `as any` casts; confirmed neither package's compiled output actually calls the removed errorAndThrow, so this is a type-only gap, not a runtime one - Dropped @colors/colors and the local isEmpty/escapeRegExp/ isPlainObject/truncateString/memoize shims in lib/utils, now that the appium peerDependency bump makes their appium/support equivalents available (both were carrying "replace once this driver declares that minimum" TODOs) - Replaced the pem devDependency in the Safari SSL e2e test with a direct `openssl req -x509` invocation - Bumped commander to v15 BREAKING CHANGE: requires Appium >=4.0.0-beta.0 and drops Node 20 support (minimum is now ^22.22.2 || ^24.15.0 || >=26.0.0), matching Appium 4's own minimum supported Node engine. The appium/app/reset endpoint is also gone; use the corresponding 'mobile:' extensions to manage app state instead. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
…tart resetProperties() previously set only the legacy pageLoadMs/implicitWaitMs fields, leaving BaseDriver's own scriptTimeoutMs/pageLoadTimeoutMs (now exposed by GET /timeouts on Appium 4) at their stock 30000/300000 defaults instead of the values that actually drive the browser session. Route through setImplicitWait/setPageLoadTimeout/setScriptTimeout instead so both stay in sync from the start, and add a regression test. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
The recorder waited a fixed 3 minutes for xctrace/instruments to exit after SIGINT. Large traces (e.g. CPU Profiler, all processes) can take longer to finalize, and on timeout the process is now killed and the trace deleted, so the caller got nothing back. Let the caller raise the limit; the default stays at 3 minutes. Co-authored-by: Cassian433 <262026629+Cassian433@users.noreply.github.com>
* fix: stop the running screen recording when forceRestart is set * fix: keep the ffmpeg process until stop succeeds so a failed stop can be retried --------- Co-authored-by: Cassian433 <262026629+Cassian433@users.noreply.github.com>
* fix: parse unquoted and RFC 5987 Content-Disposition filenames * refactor: use filenameFromContentDisposition from base-driver --------- Co-authored-by: Cassian433 <262026629+Cassian433@users.noreply.github.com>
BREAKING CHANGE: Removed the launchApp command and the POST /appium/app/launch route. Use mobile: launchApp instead. BREAKING CHANGE: Removed the closeApp command and the POST /appium/app/close route. Use mobile: terminateApp instead. BREAKING CHANGE: Removed the asyncScriptTimeout command and the POST /timeouts/async_script route. Use setScriptTimeout instead. BREAKING CHANGE: Removed the no-op getSafariIsIphone driver method. BREAKING CHANGE: Removed the no-op getSafariDeviceSize driver method. BREAKING CHANGE: Removed the no-op getSafariIsNotched driver method. BREAKING CHANGE: Removed the no-op getExtraTranslateWebCoordsOffset driver method. BREAKING CHANGE: Removed the no-op getExtraNativeWebTapOffset driver method. BREAKING CHANGE: Removed the deprecated hasElementId driver property. BREAKING CHANGE: Removed the nativeWebTapTabBarVisibility setting. It had no effect, and setting it is no longer accepted. BREAKING CHANGE: Removed the nativeWebTapSmartAppBannerVisibility setting. It had no effect, and setting it is no longer accepted. BREAKING CHANGE: Removed the safariTabBarPosition setting. It had no effect, and setting it is no longer accepted.
* fix: retain and clean up screen recorder after stop failure (cherry picked from commit 1a32f67) * test: adapt beta recorder retry regression to kill escalation
BREAKING CHANGE: Removed the undocumented XCUITestDriver.executeAtomAsync method. It had no internal callers and was not exposed as a command. Driver subclasses that call it should use this.remote.executeAtomAsync(atom, args, this.curWebFrames) followed by this.waitForAtom(promise) instead.
#3041) BREAKING CHANGE: Removed the asyncWaitMs driver property. Use scriptTimeoutMs (base driver) instead. BREAKING CHANGE: GET /timeouts reports script: null for a new session, not 0. BREAKING CHANGE: execute now uses the script timeout that the client sets (before: ignored). A client that sends 0, or a value shorter than its scripts, now gets script timeout.
* feat: add execute method for retrieving pressable buttons * chore: fix route name * docs: update description
…m unzip (#3080) BREAKING CHANGE: the bsdtar utility is no longer used and the APPIUM_PREFER_SYSTEM_UNZIP environment variable has no effect. Non-UTF-8-flagged zip file names are decoded as utf8; cp437 is not supported.
…(tunnel preferred, usbmux fallback) (#3082) BREAKING CHANGE: appium-ios-device is no longer used. BREAKING CHANGE: The lockdown simulate-location service is gone, so setting the geolocation on real devices below iOS/tvOS 17 is not supported. BREAKING CHANGE: The appium:safariSocketChunkSize capability has been removed. BREAKING CHANGE: The appium:safariWebInspectorMaxFrameLength no longer has an effect on real devices. BREAKING CHANGE: mobile: killApp always returns false on real devices below iOS/tvOS 17 (and on Linux/Windows hosts without a tunnel), because the instruments service it used is gone.
…exts (#3090) * fix: stop stale remote debugger listeners when switching webview contexts * chore(deps): bump appium-remote-debugger to ^18.0.3 * refactor: split setContext into private helpers
…ks removal for real devices (#3079)
* fix: remove perf recording temp files when start or upload fails * style: format performance spec
…rce excludedAttributes (#3078)
* fix: align default screenshotQuality setting with WDA * refactor: make WDA the source of truth for its settings
Restore beta's versions of files where the rebase of duplicated master backports regressed them (recordscreen, types, implicit-wait, driver command table) and fix a duplicated test declaration.
KazuCocoa
left a comment
KazuCocoa
left a comment
There was a problem hiding this comment.
Three reproducible timeout/compatibility issues to address before the major release. Reproduced against d9d43f9 with Appium 4.0.0-beta.7 (within the declared peer dependency range); device operations were stubbed for session initialization.
| ): Promise<DefaultCreateSessionResult<XCUITestDriverConstraints>> { | ||
| try { | ||
| const [sessionId, initialCaps] = await super.createSession(w3cCaps1, w3cCaps2, w3cCaps3, driverData); | ||
| const [sessionId, initialCaps] = await super.createSession(w3cCapabilities); |
There was a problem hiding this comment.
[P2] Synchronize the initial page-load timeout before starting the backend
Appium 4's BaseDriver.createSession() assigns the initial timeouts directly to its properties rather than calling the overridden setters. Creating a session with timeouts: {pageLoad: 45000} therefore makes getTimeouts().pageLoad return 45000, but this driver's pageLoadMs remains 6000. getNewRemoteDebugger() and mobileStartAutomationSession() both consume the latter, so the configured page-load budget is not actually used. I reproduced this by stubbing device startup and calling the real createSession(). Please synchronize pageLoadMs from pageLoadTimeoutMs after the superclass call and before start(), and cover an initial timeouts capability rather than only construction defaults and later timeout commands.
| override setImplicitWait(ms: TimeoutValue): void { | ||
| super.setImplicitWait(ms); | ||
| // the automation session needs a number: null (no limit in BaseDriver) gives 0 | ||
| this.syncAutomationSessionTimeout('implicitWaitTimeoutMs', ms ?? 0); |
There was a problem hiding this comment.
[P2] Preserve unlimited implicit waits in the automation backend
BaseDriver treats implicit: null as an unlimited wait, but ms ?? 0 turns it into a zero-wait lookup for an active AutomationSession. The same conversion also occurs in mobileStartAutomationSession(). With a stubbed evaluator that returns no element on its first call and an element on its second, the actual remote-debugger findElements() implementation returns [] after one attempt with this conversion; using Infinity finds the element on the second attempt. Please preserve unlimited-wait semantics in both initialization and updates. The existing tests currently assert zero, so they lock in the mismatch between GET /timeouts and actual lookup behavior.
| }, | ||
| "peerDependencies": { | ||
| "appium": "^3.0.0-rc.2" | ||
| "appium": "^4.0.0-beta.6" |
There was a problem hiding this comment.
[P2] Restore or explicitly remove the remaining legacy implicit-wait route
After switching to Appium 4, POST /session/:sessionId/timeouts/implicit_wait is still registered in lib/method-map.ts with command implicitWait, but neither this driver nor the new BaseDriver implements that method. On this head, typeof driver.implicitWait is undefined, and driver.executeCommand('implicitWait', 1000) rejects with NotYetImplementedError. If this deprecated route is being retained, add a bridge that validates the value and calls setImplicitWait(); if its removal is intentional for the major release, remove the route and its documentation alongside the other deleted APIs.
|
🎉 This PR is included in version 13.0.0 🎉 The release is available on GitHub release Your semantic-release bot 📦🚀 |
CI-only draft: runs the checks on the rebased
beta(see #3110). Do not merge;masteris fast-forwarded with a push instead. Will be closed afterwards.